crucible

Consolidated pre-push review gate — runs the full reviewer set (temper, delve, red-team, plus siege/inquisitor when the diff warrants) as one gate and emits a single PASS/BLOCKED verdict. Use when you want the *complete* pre-push bar in one call — "gate this before I push", "run the full review gate", "run all the reviewers", "warden". NOT for a single-reviewer pass: a fresh-eyes code review stays with temper, red-teaming a design/plan stays with quality-gate, a bug-hunt on a diff stays with delve.

Use when starting any feature development, building new functionality, implementing a design, or going from idea to working code. Triggers on "build", "implement", "add feature", or any task requiring design-through-execution.

Iterative red-teaming of any artifact (design docs, plans, code, hypotheses, mockups). Loops until clean or stagnation. Invoked by artifact-producing skills or their parent orchestrator.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup

Eval-only skill for measuring skill routing accuracy. Not invoked directly — contains selection evals that test whether the agent picks the correct skill for a given prompt.

Iteratively review code changes for production readiness through fresh-eyes review loops. Use when completing tasks, implementing major features, or reviewing a change for production readiness — including when the user says "review this PR", "review my changes", "code review", "check the diff", or "is this ready to ship". Works on PRs from any forge (GitHub, GitLab, Bitbucket, self-hosted) or on raw git SHA ranges.

Tour of the Crucible workshop — the headline orchestrators users invoke directly. Use when the user asks "what skills are available?", "where do I start?", "what should I use for X?", "give me a tour of Crucible", "what are the main commands?", or any onboarding-style question. Also use after onboarding a new user or when someone needs to pick the right tool for a specific task. This is the on-demand catalog tour; the silent conversation-start activation protocol itself is getting-started.

Security audit of design docs, implementation plans, and code. Dispatches 6 parallel Opus agents across attacker perspectives, iterates until zero Critical + zero High findings, and maintains a persistent threat model. Triggers on 'siege', 'security audit', 'security review', 'threat model', or when audit detects security-relevant surfaces.

Standalone instance-bug reviewer — runs a parallel finder fan-out + verify gate over a diff or a path and prints ranked, verified findings. Use when the user says "delve", "find bugs in this diff", "review this for bugs", "scan this file/subsystem for defects", "instance-bug sweep", or wants concrete reproducible defects (not a merge verdict, not systemic health). Works on a PR id, a base..head range, or a path, on any forge (GitHub, GitLab, Bitbucket, self-hosted).

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

The Book of Grudges — cross-session bug graveyard. Every fixed bug is recorded as a structured "grudge"; before touching code, skills query the grudgebook for the files in scope and surface past regressions as forced "DO NOT REPEAT" context. Read mode (pre-flight) and write mode (on bug resolution / fix(*) PR). Machine-local, per-repo, never committed. Triggers on /grudge, "check grudges", "record a grudge", "any past bugs here", "regression oracle", "bug graveyard".

Standalone codebase investigation. Produces a layered Investigation Brief with core findings (structure, patterns, scope, prior art) plus optional depth modules. Dispatches parallel scouts, synthesizes findings, and feeds cartographer. Use before a specific task requiring codebase understanding — task-scoped to the surface that task touches (for first-time whole-repo onboarding with cross-repo topology, use project-init, which bootstraps the cartographer data recon consults).

Internal temper-eval harness — not auto-routed; invoke explicitly via /temper-eval-calibrate. Wrapper skill that runs k iterations of (stage → collect-behavior → score) for

Internal temper-eval harness — not auto-routed; invoke explicitly via /temper-eval-collect. Live-dispatch phase of temper eval harness. Reads stage-manifest.json from a pre-staged dispatch dir; fans Task-tool reviewer dispatches in parallel (max 6); writes per-seq result files; exits. Single bounded session. Pairs with `python -m skills.temper.evals.run_evals stage` and `score`.

Adversarial review of code subsystems or non-code artifacts (design docs, plans, concepts) through parallel analytical lenses. Triggers on 'audit', 'review subsystem', 'audit this design', 'review this plan', 'audit concept', 'check the save system', 'examine the UI code', or any task requesting adversarial review of existing artifacts.

Multi-model consensus for high-stakes quality decisions. Opt-in MCP-based system that dispatches prompts to multiple LLM providers in parallel and synthesizes their responses. Used by quality-gate (stagnation judge, periodic red-team) and design (Challenger).

Use when implementing Unity UI Toolkit code from a mockup, HTML reference, screenshot, or visual spec. Triggers on "implement this mockup", "translate to USS", "build this UI", "match the mock", "UI looks wrong", "fix the layout", "spacing is off", "doesn't look like the design", or any task turning a visual reference into Unity UI Toolkit USS/C# code.

Use when you have a spec or requirements for a multi-step task, before touching code

Enforces the Detect → Fetch → Implement → Cite protocol when implementing against external frameworks or libraries. Invoke when a change touches an external API surface and the edit exceeds the triviality threshold, so implementations come from current official docs rather than stale training-data recall.

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - creates isolated git worktrees with smart directory selection and safety verification

Use after completing an implementation to find unknown failure modes. Reads the implementation diff and writes up to 5 tests designed to make it break — per-task, scoped to that single implementation (for a fully-assembled multi-component feature, use inquisitor to hunt cross-component bugs). Triggers on 'break it', 'adversarial test', 'stress test implementation', 'find weaknesses', or any task seeking to expose unknown failure modes in a single implementation.

Use when starting any conversation - establishes how to find and use skills, requiring skill activation before ANY response including clarifying questions. This is the conversation-start bootstrap protocol, not the user-facing catalog tour (for "what skills exist / which one for X", use workshop).

Use when onboarding to an unfamiliar codebase and want full structural context before the first real task. Deep-scans the repo and discovers cross-repo topology — whole-repo first-time bootstrap, not task-scoped investigation (use recon when you already have a specific task; project-init bootstraps the cartographer data recon consults).

Use when verifying Unity UI matches a mockup, after implementing UI from a visual reference, when a user shares a screenshot showing UI drift, or when checking theming compliance. Produces a drift/delta report; to implement or fix UI from a mockup, use mock-to-unity. Triggers on "verify", "compare to mock", "does this match", "check the UI", "check theming", "colors are off", "screenshot shows wrong", "visual bug", or completing any UI implementation task.

Use when a full feature is assembled and you want to hunt cross-component bugs before final quality gate. Runs 5 parallel adversarial dimensions against the complete implementation diff. Triggers on 'inquisitor', 'hunt bugs', 'cross-component test', 'find integration issues', or automatically in build pipeline Phase 4.

Ecosystem-appropriate dependency vulnerability audit. Walks the project for package.json, Cargo.toml, requirements.txt, pyproject.toml manifests and runs npm/cargo/pip-audit. Produces a structured supply-chain signal with normalized severities. Used by build alongside quality-gate; can also be invoked standalone. Triggers on /dependency-audit, 'audit dependencies', 'scan vulnerabilities', 'check CVEs'.

Audits all crucible skills for overlap, staleness, broken references, and quality. Quick scan or full evaluation modes.

Use when a significant task completes and you need a retrospective, when starting a new task and want to consult past lessons, or when 10+ retrospectives have accumulated and skill improvements should be proposed

Use when exploring unfamiliar code and want to persist what you learn, when starting a task and want to consult known codebase structure, or when collaborators need module-specific context for implementation or review

Use when you need adversarial review of any artifact — design docs, implementation plans, code, PRs, or documentation. Iterates until clean or stagnation.

Reconcile the Crucible calibration ledger — walk merged fix/hotfix branches to falsify the originating gating-verdicts, compute per-skill Brier calibration scores, and append a falsification log. Triggers on "/calibration-reconcile", "reconcile ledger", "reconcile calibration", "falsify verdicts", "brier score", "calibration reconcile", "compute brier".

Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation

Audit existing tests for staleness, needed updates, or removal after code changes. Use after any code modification to verify test suite alignment. Triggers on 'audit tests', 'check test alignment', 'stale tests', 'test health', or any task verifying tests match changed code. Technology-agnostic — works with any language and test framework.

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

Use when merging a PR after implementation is complete - verifies CI, runs local tests, checks repo safety, and monitors post-merge health. Triggers on 'merge pr', 'merge this', 'land this PR', or any task executing a PR merge.

Render the Crucible calibration ledger weekly report — the honest "Crucible caught N silent bugs" headline, verdict breakdown, per-skill severity rates, and the inflation detector. Triggers on "/ledger", "weekly report", "weekly ledger", "caught N", "quality ledger", "calibration report", "render the ledger".

Read or update the per-repo arc-state file (docs/compass.md). Tracks current arc, last meaningful commit, open loops, next move, and don't-forget items. Auto-maintained by build, merge-pr, finish; read by getting-started. Use "compass read" to inspect project state, "compass update" to set a field, "compass doctor" to validate, "compass compress" when at cap. Triggers on "compass", "current arc", "project state", "what am I working on", "arc state", "where was I".

Write a handoff doc for the next session — either continuing the current arc, or proposing the natural next pickup if the current arc is wrapping up.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Autonomous migration planning and execution. Takes a migration target (framework upgrade, API version bump, dependency major version, deprecation removal) and produces a phased migration plan with compatibility verification, then optionally executes via build's refactor mode. Triggers on /migrate, 'migration plan', 'upgrade X from Y to Z', 'remove deprecated', 'major version bump'.